Performance degradation that occurs when multiple inference requests compete for the same GPU's memory and compute resources.